Smart Viruses - When Good AI Goes Bad

As with any technology, not all uses of AI will be for the better. The existence of “smart” programs that evolve and change on their own implies that “smart” viruses that evolve to avoid detection, are not far behind.

A virus is a self-replicating computer program that spreads itself into other programs, often without the knowledge or consent of the computer owner. Worms and Trojan horses are considered types of viruses. A worm is essentially a virus that does not need to be a part of another program to reproduce. A Trojan is a virus that requires the user to execute it, and it usually “disguises” itself as a legitimate program in order to accomplish this.

In May of 2005, a virus called Zellome was found to use genetic algorithms to decrypt data on a target computer. While there were easier, simpler ways to accomplish the same thing, the use of Artificial Intelligence in a computer virus opened a new proverbial can of worms—how can genetic algorithms be used in viruses?

To some extent, virus infections already exhibit processes similar to genetic algorithms. As people find ways to eliminate viruses, they weed out the weaker viruses and leave the stronger ones to continue. Variations of viruses are also similar to mutations, another factor in genetic algorithms. However, these minor “mutations” cannot change what the virus does, only what it looks like. An intelligent virus would need to be able change its functionality as well as its appearance.

Once this happens, virus checkers will need to evolve with the viruses. Most antivirus software tracks patterns in virus activity, so if the viruses are able to change what they do, the software will not be able to find them. As the viruses grow more lifelike, they could search for other viruses in the infected system with which they can swap code and create new offspring. These viruses would be fully autonomous—they could survive without human intervention and despite extermination attempts. These viruses would breed so they could spread quickly and do the most damage.

Some people will use the technology to create malicious programs—this is somewhat inevitable, because as long as technology exists, someone will want to use it for harm. As many of these exist, however, there will be more positive uses for them in the near and foreseeable future.

See also: Genetic Algorithms, Modern Luddites, The Great Mediator